JP2019514448A - System and method for tracking patient movement - Google Patents

Abstract

A portable medical device is provided. The portable medical device includes at least one sensor configured to obtain sensor data describing patient movement and at least one processor coupled to the at least one sensor. At least one processor is configured to detect patient motion from sensor data and to classify patient motion.

Description

  RELATED APPLICATIONS This application claims priority under United States Provisional Patent Application No. 62 / 316,196, filed March 31, 2016, under section 119 (e) of US Patent Act. All subject matter described in the above application is hereby incorporated by reference in its entirety as if fully described herein.

  The present disclosure relates to systems and methods for tracking patient movement.

  Medical outcomes in the clinical setting can be improved by quickly detecting events that indicate some change in the patient's health status. In clinical settings, such as in-hospital inpatient areas, the patient's various vital signs (eg, blood pressure, pulse, respiration rate) are monitored to detect changes. Monitored vital signs are often sent to a central station that responds to any changes that the health care provider evaluates the vital signs and justifies further investigation.

  Some portable medical devices include components that can track patient movement. These components may include, for example, one or more inertial measurement units (IMUs) such as accelerometers, gyroscopes and magnetometers. At the hospital site, detailed and continuous real-time tracking of patient movement and patient orientation with portable medical devices improves patient care by addressing various situations that can not be addressed through conventional techniques. obtain. In various implementations described in this disclosure, the motion parameters include orientation parameters. Thus, unless explicitly stated otherwise, reference to the movement of the patient's body or a part of the patient's body includes the orientation in space of the patient's body or a part of the patient's body.

  Thus, in accordance with some examples of the present disclosure, a medical device is provided that includes an IMU. The medical device is configured to monitor the patient for a particular type of movement and respond to the detection of several types of movement by taking actions related to the type of movement, the result of the type of patient movement Designed to prevent or deal with it. The type of patient movement that is configured to detect and cope in some instances may include seizures, tremors, coughing, loss of directional movement, movement associated with recovery of consciousness, falling, crazy walking, snoring and fainting including. In some examples, the medical device is configured to detect these and other types of patient motion based on a baseline period or training (ie, patient specific motion recognition training) period. The motion sensor / orientation sensor is thereby attached to the appropriate part of the patient's body and the patient is instructed to perform a series of predetermined movements or actions. For example, such baselines and / or training may be recorded during initial fitting of the medical device to a patient. The data may be improved over time as the device learns the behavior of the patient's movement during use. Alternatively, or additionally, the medical device according to some examples does not refer to specific baseline information or training information, but instead, of the force acting on the patient caused by patient movement. The type of movement of the patient is detected with reference to more specific but specific characteristics of the type of movement of the patient, such as size and / or frequency of movement of the patient.

  In at least one example, a portable medical device for monitoring patient movement is provided. The portable medical device includes a plurality of motion sensors and at least one processor. The plurality of motion sensors are located at one or more anatomical locations on the patient's body and are configured to detect a plurality of motion parameters corresponding to the motion of a region of the patient's body. At least one processor is communicatively coupled to the plurality of motion sensors, receives a plurality of motion parameters corresponding to the motion of the patient's body part, stores the plurality of motion parameters in the data store, and stores the plurality in the data store The plurality of motion parameters are processed to determine motion of a region of the patient's body, and the motion of the region of the patient's body is classified based on the plurality of predetermined motion sentence features. The plurality of predetermined motion sentence features include at least one predetermined motion primitive to initiate one or more notification actions based on the type.

  In the portable medical device, the plurality of motion parameters may include at least one orientation parameter that corresponds to the orientation of a region of the patient's body. The plurality of motion sensors may include at least one of an accelerometer, a gyroscope, and a magnetometer. The plurality of predetermined motion sentence features are defined by a plurality of predetermined motion primitives, at least one predetermined motion modifier, at least one predetermined motion object, and at least one predetermined motion object. The motion sentence may include at least one predetermined series of motion sentences. At least one processor is configured to receive first sensor data obtained from a first motion sensor attached to a first position and a second motion data obtained from a second motion sensor attached to a second position. It may be configured to classify patient movement with reference to sensor data. One or more anatomical locations on the patient's body may be one of the patient's body's head, chest, legs, neck, shoulders, elbows, knees, wrists, jaws, forearms, biceps, ankles and feet Or may include more than one. The first position may include an anatomical position on the patient, and the second position may include the position of a physical object other than the patient. Physical objects may include at least one of a bed and a wheelchair. The at least one processor may be configured to classify motion based on one or more motion detection rules derived from a pre-collected database of motion information.

  In a portable medical device, at least one processor classifies motion using a motion recognition process, wherein predetermined movements measured from a plurality of patients and patient-specific movements derived during a baseline period are used. May be configured to train the motion recognition process using at least one of Patient-specific movements derived during the baseline period may be recorded during the sleep period and / or during the 6 minute walk test period. The one or more notification actions may include informing the caregiver of the patient's movement. The one or more notification actions may include alerting the patient based on the patient's movement. The at least one processor may be configured to identify another medical device remote from the portable medical device proximate to the patient and instruct the medical device to emit an alert. At least one processor may be configured to identify the current time of day and to perform an action associated with the current time of day. The at least one processor may be configured to classify the patient's movement as at least one of a loss of direction, falling, falling, beating, fainting, seizures, tremors and coughing. The portable medical device may further comprise a wearable defibrillator. The portable medical device may further comprise a mobile cardiac monitoring device.

  In another example, a system for monitoring patient movement is provided. The system includes a cardiac monitoring device. The cardiac monitoring device includes a plurality of motion sensors and at least one processor. The plurality of motion sensors are configured to be located at one or more anatomical locations on the patient's body to detect a plurality of motion parameters corresponding to the motion of a portion of the patient's body. At least one processor is communicatively coupled to the plurality of motion sensors. The at least one processor is configured to transmit to the remote server a plurality of motion parameters corresponding to the motion of the patient's body part. The remote server receives the plurality of motion parameters corresponding to the motion of the patient's body part, stores the plurality of motion parameters in the data store, and processes the plurality of motion parameters stored in the data store to obtain the patient's body. Determine the movement of the body part, classify the movement of the body part of the patient based on a plurality of predetermined movement sentence features including at least one predetermined movement primitive, and one or more notifications based on the type Configured to initiate an action. The system may further include at least one client device in communication with the remote server. The at least one client device is configured to notify at least one of the caregiver and the patient of the patient's movement based on the one or more notification actions.

  In one example, a portable medical device is provided. The portable medical device includes at least one sensor and at least one processor. At least one sensor is configured to obtain sensor data describing a patient's movement. At least one processor is connected to at least one sensor. At least one processor detects motion of the patient from the sensor data, eg, whether the measured and detected motion features are derived from a pre-collected database of motions measured from a plurality of patients or other users Classify the patient's movement using template matching known to those skilled in the art, as compared to a template derived from a patient-specific template derived from a patient-specific baseline period or training period, or It is configured to execute a notification component associated with one predetermined template.

  In a portable medical device, at least one sensor may include at least one of an accelerometer, a gyroscope and a magnetometer. At least one processor may be configured to classify patient motion as at least one motion sequence including at least one motion primitive. The at least one sensor may include a plurality of sensors. At least one processor is configured to: first sensor data acquired from a first sensor attached to a first position; and second sensor data acquired from a second sensor attached to a second position And may be configured to classify the movement of the patient. The first position and the second position may be an anatomical position on the patient. The first position and the second position may be separated by the patient's joint. The joint may be at least one of the knee and elbow of the patient. The first position may be an anatomical position on the patient and the second position may be the position of a physical object other than the patient. Physical objects may include at least one of a bed and a wheelchair. Baseline information and / or training information (see, eg, FIG. 5 below) may be recorded during sleep and / or during a 6 minute walk test. Sensor data describing patient movement can not indicate movement (including orientation).

  In the portable medical device, the notification component may be configured to notify a caregiver of patient movement. The notification component may be configured to alert the patient. The at least one processor may be associated with the patient's condition and may be configured to identify a medical device proximate to the patient and instruct the medical device to emit an alert. The condition may be associated with the patient's movement. The at least one processor may be configured to perform an action related to the age of the patient. At least one processor may be configured to identify the current time of day and to perform an action associated with the current time of day. The portable medical device may be configurable between daytime mode and nighttime mode. The at least one processor may be configured to identify the current mode as either a daytime mode or a nighttime mode, and to perform an action associated with the current mode. The at least one processor may be further configured to prompt the patient to perform a series of movements and to record a template of the patient performing each of the series of movements.

  In another example, a portable medical device is provided. The portable medical device includes at least one inertial measurement unit and at least one processor coupled to the at least one inertial measurement unit. The portable medical device is configured to at least partially detect a fall of the patient by identifying a series of measurements. The series of measurements include a first acceleration below the first threshold, a second acceleration greater than the second threshold, and a third acceleration within the range of the threshold.

In portable medical devices, the first threshold may be equal to 0.5G ₀ and the second threshold may be equal to 8G _0, and the range of these thresholds is [0.95 G ₀ , 1. May be equal to 05G ₀ ]. The at least one processor may be configured to calculate a confidence metric and to detect a fall of the patient if the confidence metric exceeds a fourth threshold. The at least one processor may be configured to operate when a fall of the patient is detected. The action includes at least one of prompting the patient to confirm a fall, notifying bystanders of the fall, and notifying hospital personnel of the fall.

  In another example, a hospital-based portable defibrillator is provided. The hospital-based portable defibrillator comprises at least one adhesive electrode configured to obtain data describing a patient, at least one inertial measurement unit, at least one adhesive electrode and at least one And at least one processor coupled to the inertial measurement unit. The at least one processor is configured to monitor patient movement via the at least one inertial measurement unit.

  In a hospital-based portable defibrillator, the at least one processor may be configured to classify the patient's movement as at least one of a loss of directional movement, falling, beating and fainting. In some implementations, the at least one processor derives patient motion from a pre-collected database of motions measured from multiple patients or other users, or during a patient-specific baseline or training period. The movement of the patient may be configured to at least partially classify by comparison with a template derived from a patient specific template derived. For example, patient specific templates may be recorded during a six minute walk test. The at least one processor may be configured to classify the patient's movement as at least one of seizure, tremor and cough. The at least one processor may be configured to at least partially classify the patient's movement by comparing the magnitude of the patient's movement with the first threshold and the frequency of the patient's movement with the second threshold. . The at least one processor classifies the patient's movement as at least one type of movement selected from the group comprising directionality movement, fall, rattling, fainting, fainting, seizures, tremors and coughing, and the patient's movement It may be configured to perform at least one action in response to the classification. At least one action is to prompt the patient to confirm that the patient's movement is of at least one type of movement, to notify bystanders of the patient's movement, to notify hospital personnel about the patient's movement May include at least one of the following. The at least one processor may be configured to calculate a distance of the patient from the hospital bed and to perform at least one action in response to the distance being above a threshold. At least one action includes at least one of prompting the patient to respond, informing the bystanders of the patient's movement, and informing hospital personnel of the patient's movement.

  In another example, a system of medical devices is provided. The system includes a hospital based portable defibrillator and a hospital bed. The hospital based portable defibrillator comprises at least one adhesive electrode, at least one inertial measurement unit, and at least one processor coupled to the at least one adhesive electrode and the at least one inertial measurement unit. Including. The at least one processor is configured to monitor patient movement via the at least one inertial measurement unit. The hospital bed may include one or more inertial measurement units and one or more processors coupled to the one or more inertial measurement units. The one or more processors are configured to monitor the patient's movement via the one or more inertial measurement units.

  In the system of medical devices. The hospital based portable defibrillator may further include a first network interface. The hospital bed may include a second network interface configured to communicate with the first network interface. The one or more processors may be configured to transmit patient motion data to the at least one processor. The at least one processor may be configured to detect a patient at least partially away from the hospital bed by processing patient movement data.

  In another example, a hospital-based portable defibrillator is provided. The hospital-based portable defibrillator comprises at least one adhesive electrode configured to obtain data describing a patient, at least one inertial measurement unit, at least one adhesive electrode and at least one And at least one processor coupled to the inertial measurement unit. At least one processor is configured to monitor patient movement during sleep via the at least one inertial measurement unit.

  In a hospital-based portable defibrillator, at least one processor may be configured to classify the patient's movement as either excessive or not excessive. At least one processor may derive patient movements from a pre-collected database of movements measured from a plurality of patients or other users, or patients derived during patient-specific baseline or training periods It may be configured to at least partially classify the patient's movement by comparison with a template derived from a unique template. For example, a patient = specific template may be recorded during standard patient sleep. The at least one processor may be configured to at least partially classify the patient's movement by comparing the magnitude of the patient's movement with the first threshold and the frequency of the patient's movement with the second threshold. .

  In a hospital-based portable defibrillator, the data describing the patient may include electrocardiogram data. The at least one inertial measurement unit may be configured to obtain a heart sound of the patient. The at least one processor may be configured to analyze patient motion and / or electrocardiogram data and heart sounds to determine the condition of the patient.

  The hospital based portable defibrillator may further include a network interface coupled to the at least one processor. The data describing the patient may include electrocardiogram data. The at least one processor may be configured to transmit patient motion and electrocardiogram data to different medical monitoring devices.

  In a hospital-based portable defibrillator, at least one processor may be configured to change the arrhythmia treatment protocol in response to the patient's movement being classified as excessive. The at least one processor may be configured to change the arrhythmia treatment protocol by adding haptic alerts and alerts for hospital personnel to the beginning of the predefined arrhythmia treatment protocol.

  Further aspects, examples and advantages of these aspects and examples are described in detail below. Moreover, both the above information and the following modes for carrying out the invention are merely exemplary embodiments of the various aspects and features, and only the aspects and embodiments described in the claims. It should be understood that it is intended to provide an overview or framework for understanding the nature and features. Any of the examples or features disclosed herein may be combined with any other example or feature. References to different examples are not necessarily mutually exclusive, and are intended to indicate that the particular features, structures or characteristics described in connection with the examples may be included in at least one example. Such terms appearing herein are not necessarily all referring to the same example.

  Various aspects of at least one example are described below with reference to the accompanying drawings. The drawings are not intended to be drawn to scale. The figures are included to provide illustration and a further understanding of the various aspects and examples, and are incorporated into and constitute a part of this specification, but any specific example. It is not intended as a definition of limitation. The drawings, together with the remainder of the specification, serve to explain the principles and operation of the described and claimed aspects and examples. In the figures, each identical or nearly identical component that is illustrated in various figures is represented by a like numeral. For purposes of clarity, not every component may be shown in every figure.

FIG. 7 is a schematic view of a sensor arrangement including a plurality of motion sensors / orientation sensors disposed in proximity to a patient and in communication with a motion tracking unit according to an example of the present disclosure.

FIG. 1 is a schematic view of a portable medical device according to an example of the present disclosure.

FIG. 7 is a schematic view of a medical device controller according to an example of the present disclosure.

FIG. 5 is a schematic view of the motion information data store of the medical device controller of FIG. 3 according to an example of the present disclosure.

FIG. 6 is a flow diagram illustrating a baseline / training process configured to be performed when a motion recognition component or motion classification component operates in a baseline mode / training mode, according to an example of the present disclosure.

FIG. 6 is a flow diagram illustrating a method of detecting patient motion and classifying the detected motion as corresponding to a motion primitive or a series of motion primitives according to an example of the present disclosure.

FIG. 7 is a graph of the patient's force vector versus time during a patient's unsupported fall to the ground.

FIG. 7 is a graph of patient motion vector versus time during a patient's unsupported fall into the ground and after a collision of the patient with the ground.

FIG. 7 is a schematic view of the phase of patient movement during and after patient unsupported fall to the ground.

FIG. 16 is a graph showing the vector sum of patient motion during the onset of patient tremor.

FIG. 7 is a graph showing the vector sum of patient motion during a patient's cough episode.

FIG. 7 is a graph showing vector sums of patient motion during patient seizure episodes.

FIG. 1 is a schematic diagram of an example of a distributed computer system in a hospital environment according to an example of the present disclosure.

  The example described herein includes a portable medical device having at least one sensor used to obtain motion parameters including data describing patient motion. As described in further detail below, in various implementations, the motion parameters include patient orientation parameters. Thus, in the present disclosure, reference to movement of the patient's body or movement of a site of the patient's body includes the orientation of the body of the patient or the body of the patient in space. The sensor may be attached to the garment being worn by the patient by some removable connection means, for example by means of a band or temporary adhesive compatible with human skin or clothes, or to the patient himself It can be attached. Regardless of which part of the patient (e.g. head, arms, legs, chest) the sensor is attached, the movement (including orientation) of that part of the patient is measured. The portable medical device uses the acquired data to detect patient movement and, in some instances, classify patient movement. Data corresponding to the measured movement is collected and processed to estimate movement (including orientation in space) at that particular sensor position. The motion or orientation measurement may be used to detect a series of one or more predetermined motion or orientation types ("motion primitives"), and also to use motion primitives based on the features of the motion waveform. It is processed to classify it as corresponding to one or more predetermined motion primitives. Some examples of movement primitives are "Lit down", "Sit", "Stand up", "Put on foot", "Fall", "Roll", "Walking", "Run", "Turn around" ',' Bent ',' upright '.

  Additionally, the motion modifiers of the motion primitives are “fast” or “slow” to describe the speed of the motion, or “top”, “bottom”, “left”, “right” to describe the direction, etc. May exist. Additionally, there may be motion objects such as, for example, "legs" or "arms" or "body" for the position on the body where the sensor is located. Additionally, for example, "left" or "right" which describes which arm or leg is moving, or "which part of the torso, arm, leg or other body part is moving" There may be motion object modifiers (ie, motion adjectives) such as "up" or "down." In some cases, there may be an associated detected set of predefined rules for the detected sequence of motion primitives or motion modifiers. The predetermined rules constitute a motion grammar. Using the detected motion primitives and motion modifiers, and based on the motion grammar, for example, "step out", "rise from bed", "open the door", "walk fast in the corridor" or "on the ground" A motion sentence such as “fall over” can be configured.

  This motion recognition process of detection of motion primitives, motion modifiers, motion objects, motion adjectives and motion grammar structures may be implemented in a "baseline" period or training (i.e. patient specific motion recognition training) period. Thereby, the motion sensor / orientation sensor is attached to an appropriate part of the patient's body, and the patient can, for example, lie down, sit down, rise from sitting down, rise from lying down, walk etc. You are instructed to perform a defined movement or act. The system then analyzes the patient's specific movements or behaviors and uses it to improve the patient's movement recognition accuracy. Alternatively, motion recognition processing may not utilize any baseline period or training period, relying entirely on motion detection rules derived entirely from a database of motion collected prior to rule generation. Sometimes.

  Then, whether the output of motion recognition (MR) processing (eg, motion primitives, motion modifiers, motion objects, motion adjectives and motion grammar structures) pose a threat to the patient's health or safety, or If not, it can be analyzed to determine if it constitutes an early failure. Next, in response to the classification of the detection operation, a notification component (such as an alarm or notification to a healthcare provider or other caregiver) may be performed.

  The estimated movement (including the orientation in space) at that particular sensor position is a parameter related to movement, eg acceleration along one or more axes (eg x-axis, y-axis, z-axis) It may be a vector of velocity, position and rotation. Motion vectors may be associated with motion recognition processes such as time of day, location (e.g., geographical location recognized via a GPS system, or indoor location such as a bedroom, stairs, entrance aisle, outdoor) or weather, etc. May also contain information on In some embodiments, the axes of an accelerometer or other type of motion or orientation sensor may be located on or near the sternum and may be the main axis of the body, eg, the longitudinal axis (eg, the longitudinal axis, eg, acceleration). Aligned with the sensor's "z" axis), transverse axis (horizontal axis, eg, acceleration sensor's "x" axis) or sagittal axis.

  In some embodiments, motion recognition may be based on a Hidden Markov Model (HMM) that takes a series of vectors of motion parameters as input. State transition matrices can be created using HMM techniques known to those skilled in the art. In particular, a series of motion parameter estimates, motion primitives and motion sentences are modeled as a Hidden Markov Model (HMM) and have a set of state Q, output alphabet O, transition probability A, output probability B and initial state probability Π. Defined as a finite state machine variable. The current state is not observable. Instead, each state produces an output with a particular probability (B). Since the state Q and the output O are usually understood, the HMM is said to be a triple λ = (A, B, Π). Each value of the output alphabet O may be given a unique threshold and a set of coefficients. For example, the HMM will have transition probabilities stored in the HMM as a result of either database or patient specific training. The HMM indicates such an understood "fact" (i.e. motion grammar). For example, the probability of getting up shortly after a person is standing is very low. Intermediate motion primitives that can be lost in noise, or alternatively, both “rise up” and “fall over”, “bent” or “fall over” can all involve the patient's thorax in a forward rotation. There must have been any intermediate motion primitives in which the probability of transitioning to either "falling", "bending" or "falling" is higher than the probability of transitioning to "rise".

  In another embodiment, the grammatical structure of the "stand" motion sentence follows the detection or flexion of the patient's thorax, and then returns to the "stand" orientation (eg, "stand", "bent", " If standing up and “bending up” continue twice or more, the grammar sentence structure is at high risk of injury due to the patient losing balance and does the MR output pose a threat to the patient's health or safety? Or otherwise detected to mean constituting an initial failure. Next, in response to the classification of the detection operation, a notification component (such as an alarm or notification to a healthcare provider or other caregiver) may be performed.

  During algorithm development, HMMs are trained against databases of multiple motion types (eg, sitting, standing, walking, lying) by multiple patients or other subjects. In some embodiments, a patient-specific baseline period or training period is provided to improve the results of the basic approach without a patient-specific training period. For example, during a training period, normalizing motion parameters to the size of the patient, eg torso or limb length, eg, getting up from bed, rolling on bed, entering bed, getting out of bed, climbing stairs, Training a particular movement sequence by a particular patient, such as walking, performing context movement by performing multiple movements in different contexts (eg, in different rooms or beds, or at different times) The same test), intensity dependent training of the patient (e.g. if the patient performs the same action at different intensity levels) may be used for improvement.

  Other techniques known in the field of HMM classification may be utilized. For example, a discriminatory training technique that eliminates the purely statistical approach to HMM parameter estimation and instead optimizes some classification-related measures of training data. Examples include maximum mutual information (MMI), minimum classification error (MCE) and minimum phoneme error (MPE). There is also the use of the Viterbi algorithm to find the best path. Other techniques are just to retain the set of good candidates instead of keeping the best candidates, and to rank these good candidates using a better scoring function (re-scoring) . The set of candidates may be kept as either a list (N-best list approach) or a subset of the model (lattice). Re-scoring may be performed to minimize Bayesian risk.

  In some embodiments, techniques such as dynamic time warping may be utilized to minimize the impact on recognition accuracy due to the speed of motion. For example, similarities in walking patterns may be detected even if the patient is moving his legs more slowly than in a training session. The sequences are "stretched" non-linearly to match each other. This sequence alignment method is often used in conjunction with HMMs.

  In some embodiments, so-called deep learning networks can be utilized for motion recognition. Fundamentals and trends in “Deep learning: methods and applications” signal processing, co-authored by Deng Li and Yu Dong: (7 (3-4): 197-387) (2014) as described by Deng and Yu, Deep learning networks can be classified as three main classes:

  1. Deep networks for unsupervised learning or generative learning intended to capture higher order correlations of observed or visible data for pattern analysis or synthetic purposes when information about target class labels is not available . Unsupervised feature learning or unsupervised expression learning in this document refers to this category of deep networks. It is also intended to characterize the combined statistical distributions of visible data and their associated classes (if available and treated as part of visible data) when used in production mode is there. In the latter case, Bayesian rules can be used to make this type of generator network a learning discriminant network.

  2. A deep network for supervised learning intended to provide discriminatory ability directly (in many cases by characterizing the posterior distribution of classes conditioned on visible data) for the purpose of pattern classification. Targeted label data is always available in direct or indirect form in such supervised learning. They are also called discrimination deep networks.

  3. A hybrid deep network where discrimination is the goal. This determination is often aided in a salient way by using the results of generated deep networks or unsupervised deep networks. This may be achieved by better optimization and / or regularization of category (2) deep networks. This goal may also be realized if discriminant criteria for supervised learning are used to estimate the parameters in either the generated deep network of category (1) or the unsupervised deep network.

  Other common types are deep neural networks (DNN) and recursive neural networks (RNN), convolutional neural networks (CNN), restricted Boltzmann machines (RBM), deep belief networks (DBN), deep Boltzmann machines (DBM) Etc.

  In some embodiments, simpler MR techniques may be utilized, such as template matching as known to those skilled in the art. In the art, measured and detected motion features are derived from a pre-collected database of measured motion from multiple patients or other users, or derived during a patient-specific baseline or training period. The template is compared to the template derived from the patient specific template.

  Exemplary Sensor Placement FIG. 1 is a schematic view of a sensor placement 100 configured to detect patient orientation and movement at particular sensor locations on the patient's body. Data corresponding to the measured movement or orientation is collected and processed to estimate movement or orientation at that particular sensor position. The motion or orientation measurement may be used to detect a series of one or more predetermined motion or orientation types ("motion primitives"), and also to use motion primitives based on the features of the motion waveform. It is processed to classify it as corresponding to one or more predetermined motion primitives. Some examples of movement primitives are "Lit down", "Sit", "Stand up", "Put on foot", "Fall", "Roll", "Walking", "Run", "Turn around" ',' Bent ',' upright '.

  Additionally, the motion modifiers of the motion primitives are “fast” or “slow” to describe the speed of the motion, or “top”, “bottom”, “left”, “right” to describe the direction, etc. May exist. Additionally, there may be motion objects such as, for example, "legs" or "arms" or "body" for the position on the body where the sensor is located. Additionally, for example, "left" or "right" which describes which arm or leg is moving, or "which part of the torso, arm, leg or other body part is moving" There may be motion object modifiers (ie, motion adjectives) such as "up" or "down." In some cases, there may be an associated detected set of predefined rules for the detected sequence of motion primitives or motion modifiers. The predetermined rules constitute a motion grammar. Using the detected motion primitives and motion modifiers, and based on the motion grammar, for example, "step out", "rise from bed", "open the door", "walk fast in the corridor" or "on the ground" A motion sentence such as “fall over” can be configured.

  This motion recognition process of detection of motion primitives, motion modifiers, motion objects, motion adjectives and motion grammar structures may be implemented at "baseline" (i.e. motion recognition training). Thereby, the motion sensor / orientation sensor is attached to an appropriate part of the patient's body, and the patient can, for example, lie down, sit down, rise from sitting down, rise from lying down, walk etc. You are instructed to perform a defined movement or act. The system then analyzes the patient's specific movements or behaviors and uses it to improve the patient's movement recognition accuracy. Alternatively, motion recognition processing may not utilize any baseline period or training period, relying entirely on motion detection rules derived entirely from a database of motion collected prior to rule generation. Sometimes.

  Then, whether the output of motion recognition (MR) processing (eg, motion primitives, motion modifiers, motion objects, motion adjectives and motion grammar structures) pose a threat to the patient's health or safety, or If not, it can be analyzed to determine if it constitutes an early failure. Next, in response to the classification of the detection operation, a notification component (such as an alarm or notification to a healthcare provider or other caregiver) may be performed.

  The exemplary sensor arrangement 100 shown includes a plurality of motion sensors / orientation sensors 104 and a motion tracking unit 106. All of them are attached to or worn by the patient 102. The motion sensor / orientation sensor 104 of the sensor arrangement 100 may include one or more inertial motion unit sensors ("IMUs"). The IMU may include, but is not limited to, one or more of a one-dimensional accelerometer, a two-dimensional accelerometer, a three-dimensional accelerometer, a gyroscope and a magnetometer. For example, an example of the motion sensor / orientation sensor 104 may be an ANALOG DEVICES® ADIS 16362 iSensor® inertial system or an iNEMO® M1 motion sensing system manufactured by STMicroelectronics®, etc. Includes axis accelerometers and multi-axis gyroscopes (including multi-axis magnetometers). Another example of the motion sensor / orientation sensor 104 includes MMA7361 LC (measuring range ± 1.5 g, ± 6 g) which is a Freescale Semiconductor 3-axis low g ultra-small accelerometer.

  Calculations of data collected by the motion sensor / orientation sensor 104 may be performed by a microprocessor located internally to the motion sensor / orientation sensor 104 itself, or by another location micro-communicating with the motion sensor / orientation sensor 104 Executed by a processor. The microprocessor uses the data (e.g., linear acceleration, force, orientation) collected by the motion sensor / orientation sensor 104 to determine the position and orientation of the motion sensor / orientation sensor 104 in three-dimensional space.

  Generally, one or more IMUs change the position of the patient's body part by at least three axes: pitch (up / down relative to cross-section), yaw (left / right relative to sagittal plane) And roll (clockwise / counterclockwise with respect to the coronal plane, eg, backward or forward). Each of the measurements in these axes causes the one or more IMUs to generate a series of motion parameters corresponding to the region of the patient's body to which the one or more IMUs are attached. For example, for every six degrees of freedom (such as x, y and z and θx, θy and θz, etc.), the IMU may integrate the sensed acceleration in time with an estimate of gravity to calculate the current velocity. The IMU may then integrate this velocity to calculate the current position. Based on the position information, the IMU may derive one or more orientation parameters for the body part of the patient for use in the motion recognition / classification process.

  Depending on the type of sensor used for the motion sensor / orientation sensor 104, motion is detected including orientation in one or more dimensions of three-dimensional space (ie Cartesian x-direction, y-direction and z-direction) Be done. Data corresponding to this movement (whether acceleration data, orientation data or other measurement data) is detected by the motion sensor / orientation sensor 104 and provided to the motion tracking unit 106. This data may also indicate no movement (ie, the patient is not moving). Furthermore, an identifier (ID) corresponding to each sensor providing data is transmitted to the motion tracking unit in the state at the time of detection. These data are then used to classify motion. This is described in more detail below in connection with FIGS. 4 to 6.

  In the exemplary sensor arrangement 100 shown, the motion sensor / orientation sensor 104 is serially connected to one another along distinct paths terminating at a distal point of the patient 102. As shown, these various exemplary connection paths originate from motion tracking unit 106. One path connects several sensors along a first path that terminates at the patient's 102 right arm. Another pathway connects several sensors along a second pathway that terminates at the patient's 102 calvaria. Other similar pathways extend to the patient's 102 left arm and the patient's 102 right and left arms. Other paths may connect one or more sensors located at other parts of the patient's 102 body to the motion tracking unit 106, as desired. Additionally, in some instances, some of the sensors 104 may be omitted.

  The position of the individual motion sensor / orientation sensor 104 on the patient's 102 body is selected with the intention of detecting movement of the portion of the patient 102 to which the corresponding motion sensor / orientation sensor 104 is attached. In the example shown, the motion sensor / orientation sensor 104 is attached to an anatomic site of the patient's body to detect movement of specific areas of different parts of the arm, leg, head or torso, or , Generally located at an anatomical location on the patient 102, optionally separated by joints (eg, elbows, knees, wrists, necks, jaws). Anatomical locations as described herein include any location on the patient's body where one or more motion sensors may be deployed to monitor movement including orientation of the underlying region of the patient's body. obtain. For example, such anatomical locations include one or more of the patient's body head, chest, legs, neck, shoulders, elbows, knees, wrists, jaws, forearms, biceps, ankles, and feet. obtain.

  Configuring the motion sensor / orientation sensor 104 in this way is useful for detecting the patient's general motion, but the motion of all parts of the patient's 102 body, and even one part of the patient 102. It is also useful to detect movement relative to other parts of Motion sensor / orientation sensor 104 may also be configured to transmit data indicating that there is no motion of the patient's 102 site. By way of example, a motion sensor / orientation sensor 104 attached to the torso (or attached to the torso) may help to determine that the patient 102 has fallen to the ground. In another example, a motion sensor / orientation sensor attached to one or more legs may help to determine that the patient 102 is walking. Other similar examples will be apparent in light of the foregoing examples and the present disclosure.

  In another example, one or more sensors 104 of the sensor arrangement are attached to a stationary object such as a bed or chair (or even a wheelchair that exhibits uniform linear motion that can be distinguished from patient movement) . Thereby, the patient's movement is detected with respect to the known stationary object. For example, if the patient 102 wearing at least one motion sensor / orientation sensor 104 comes out of the bed to which another motion / orientation sensor 104 is attached, then this relative movement between these two sensors is It is detected. This can be particularly beneficial for notifying caregivers of high risk patient movements. For example, using such a sensor configuration, a nurse may be immediately alerted that a patient 102 at high risk of falling is in the process of exiting the bed. Thus, a nurse can intervene to assist the patient. Similarly, this configuration (and others described herein) can be used to alert the patient to alert themselves if a high risk movement is detected. Thus, the patient can wait for help and / or stop the high risk movement.

  The exemplary sensor arrangement 100 shows a plurality of motion sensors / orientation sensors 104 connected via wires to other sensors on a path and eventually to the motion tracking unit 106 Not required. It will be appreciated that in other examples, motion sensor / orientation sensor 104 may be connected to motion tracking unit 106 or a transceiver of a distributed computing system not located on the patient via a wireless connection.

  Exemplary Portable Medical Devices In some cases, the examples described herein may be worn by, attached to, or otherwise attached to a patient for an extended period of time without substantial interruption. If not, it can be connected to the patient. For example, the device may be worn by the patient, attached to the patient, or otherwise connected to the patient for a period of hours, days, weeks or longer. For example, the portable device may be configured for substantially continuous monitoring and / or treatment over a variety of time periods, including at least a 24 hour period, at least a week and at least a month. For example, such medical devices may include monitoring devices and / or treatment devices configured to continuously monitor the patient for a particular medical condition over an extended period of time. Long-term refers to, for example, 4 hours (eg, treatment and monitoring devices such as sleep apnea devices), 12 hours (eg, mobile cardiac monitoring devices, wearable defibrillator devices, etc.) treatment devices and / or monitoring Device) and a substantially continuous monitoring period over a period of 24 hours or even several days. Such devices may monitor the patient substantially continuously, apart from the period in which the patient may periodically remove the device to take a shower, re-attach, change components of the device, etc.

  FIG. 2 shows an example portable external patient wearable 102 patient medical device 200, such as LifeVest®, a wearable defibrillator available from ZOLL Medical Corporation. As shown, the medical device 200 includes a garment 210, a plurality of sensing electrodes 212, a plurality of treatment electrodes 214, a medical device controller 220, a connection pod 230, a patient interface pod 240, and a belt 222. . A motion sensor or orientation sensor (or more generally, an inertial measurement unit) 250 is also part of the medical device 200, in this case attached to the belt 222.

  The plurality of sensing electrodes 212 may be placed at various locations around the patient's body (eg, may be held by the garment 210 in place or attached to adhere to the patient's skin). As shown, sensing electrode 212 is electrically coupled to medical device controller 220 through connection pod 230. In some implementations, some of the components of wearable medical device 200 are attached to a garment 210 that may be attached to the patient's torso. For example, as shown in FIG. 2, the patient 102 may be attached to the patient with at least some of the controller 220, at least some of the sensing electrodes 212, and optionally, an adhesive, to the patient. Can be mounted on the belt 222 which is mounted by The sensing electrodes 212 and the connection pods 230 may be collected on the garment 210 or integrated into the garment 210 as shown. The sensing electrode 212 is configured to obtain a signal indicative of a patient's cardiac function. The controller 220 may thereby monitor one or more cardiac signals of the patient. Multiple treatment electrodes 214 may be electrically coupled to controller 220 through connection pod 230. The treatment electrode 214 is configured to apply one or more therapeutic defibrillation shocks to the patient's body when the controller 220 determines that such treatment is justified. The connection pod 230 may include electronic circuitry and one or more sensors (eg, motion or orientation sensors, accelerometers, etc.) configured to monitor the patient's activity.

  Wearable medical device 200 may include an optional patient interface pod 240 coupled to medical device controller 220. For example, the patient interface pod 240 may include patient interface elements such as speakers, a microphone responsive to patient input, a display, an interactive touch screen responsive to patient input, and / or physical buttons for input. In some implementations, these elements are incorporated into the housing of controller 220. The patient interface pod 240 may be wirelessly connected to the controller 220, eg, through a network interface, thus transmitting electrocardiogram data to separate and distinct medical monitoring devices. Patient interface pod 240 may take other forms and may include additional features. For example, patient interface pod 240 may be implemented on a smartphone, tablet or other portable device carried by a patient. In another example, the patient interface pod 240 can be worn around the patient's wrist as a watch, or worn around the patient's upper arm as a band. In some implementations, controller 220 may communicate certain alerts and information, and / or may respond to patient input via patient interface elements included in controller 220 and / or patient interface pod 240. The patient and / or caregiver may interact with the touch display or patient interface pod 240 to control the medical device 200.

  In addition to accelerometers that may optionally be present on the wearable medical device 200 to detect heart sounds and other activities of the patient, a motion sensor or orientation sensor 250 may also be included in the wearable medical device 200 in some instances. It is attached. In some instances, the motion sensor or orientation sensor 250 itself may be configured to detect a patient's heart sound, thus monitoring heart function and measuring the patient's physiological condition via heart sound and lung sound. It can also operate as a sensor of In some cases, these measurements are used to treat the patient (eg, defibrillation, cardiac pacing, CPR, treatment of dyspnea, cardiac disorders, etc.). Although the example in FIG. 2 shows motion sensor or orientation sensor 250 attached to belt 222, motion sensor or orientation sensor 250 may be attached to other parts of wearable medical device 200. Motion sensor or orientation sensor 250 may include at least one of a three-dimensional accelerometer and a gyroscope. This allows the orientation of the patient 102 to record baseline data or training data to populate the data structure (described below with reference to FIG. 4) and to detect and analyze the patient's 102 movement. Both can be judged. By using at least one of a three-dimensional accelerometer and a gyroscope, the wearable medical device 200 can be configured such that the torso (e.g. chest) of the patient 102 or other part of the body (e.g. It is possible to detect motion primitives or sentences that the forearm, biceps, ankles, neck) accelerates or turns. An example of a movement primitive or movement sentence that includes a change in torso movement and / or orientation may include several types of footstepping, falling, fainting and / or beating, rising from the bed, turning.

  In another example, the wearable medical device 200 is coupled to (or otherwise coupled to) a sensing electrode 212 and a motion sensor or orientation sensor (or inertial measurement unit) to monitor patient movement during sleep. May include a processor). In monitoring patient motion during sleep, the processor may use the methods and systems described herein to determine if the patient's motion is excessive. For example, during a baseline period or training period, such as the first night a patient wears the wearable medical device 200, normal sleep movement may be recorded while the patient is asleep. When using template-based motion recognition or heuristic logic-based motion recognition, a motion sentence "sleep" (ie, "beside quietly for a sufficient amount of time"), which is a feature: Based on this data acquired during the training period for those that determine the motion primitives of “to be” and the motion adjectives of “quietly” and “sufficient time”). And / or rules may be established. For example, the threshold of the angular position of the patient's torso may be, for example, 15 degrees or less, to note that the patient prefers to sleep with an additional pillow or to raise the head of the bed It may be predetermined as one or may be adjusted by the caregiver during baseline processing to suit the patient's particular situation. Similarly, based on the activity during sleep, and in some cases based on classification by the caregiver or others examining the measurements during the training period, the level of activity may for example It can be judged as "activated sleep" or "awake sleep". When HMMs or deep learning networks are used for motion recognition, recording data for that particular patient during baseline or training periods improves motion recognition accuracy. Time may also be included in the motion vector as a data element to increase accuracy. Because most individuals are more likely to be sleeping at night.

  Types of movement monitored and classified include, but are not limited to, leg, arm, head and torso movements, direction and pattern of movement. Motion thresholds during normal sleep (eg, frequency of motion, amplitude of motion) may be established and used to classify patient motion and / or trigger actions. In addition to these applications, the processor may also be configured to change the cardiac arrhythmia treatment protocol in response to classification of excessive patient motion. For example, in some implementations, a treatment protocol may associate a confidence score with a detected arrhythmia condition. In situations where the processor determines that the patient's movement is excessive (eg, the movement exceeds a predetermined threshold of the movement and / or the duration of such movement), the treatment protocol may Confidence scores that are lower than the protocol would associate if there is no motion may be associated with the detected arrhythmia condition. In some implementations, the defibrillation current to be applied to the patient can be delayed for a period of time to allow for an improvement in the confidence score. Alternatively or additionally, the altered defibrillation current may be applied to the patient based on the type or degree of excessive movement during sleep as determined by the processor. In addition, the processor may also add an arrhythmia treatment protocol by prepending one or more additional alerts, alerts and / or notifications (eg, tactile alerts to patients, alerts to hospital personnel, etc.) to the predefined arrhythmia treatment protocol. I can change it.

  Patient motion may be classified as either over or not over using the methods and systems described herein, and in some instances using thresholds. In other words, the frequency, magnitude, duration, and combinations of movements of body parts are classified in light of the movement data obtained while the patient is sleeping during the baseline or training period. obtain. For example, the magnitude of the patient's movement may be compared by the healthcare provider to a corresponding threshold set. In another example, the frequency of patient movement may be compared by the healthcare provider to the corresponding threshold set. In yet another example, the duration of the patient's movement may be compared by the health care provider to the set corresponding duration threshold. Any or all of these exemplary thresholds may trigger actions such as, for example, alerting a healthcare provider of excessive movement via a notification component if they are exceeded.

  FIG. 3 shows a schematic diagram of an example of a medical device controller 300 that receives motion sensor data or orientation sensor data and stores this data for classification in accordance with the methods and systems described herein. Optional components of the medical device controller are depicted in FIG. 3 using dashed lines. Medical device controller 300 is incorporated into motion tracking unit 106 or controller 220 in some examples.

  The medical device controller 300 includes an optional treatment delivery interface 302, a data storage 304 (including a motion information data store 330 described below with reference to FIG. 4), and an optional network interface 306. An optional user interface 308, a sensor interface 312, an optional motion recognition classification component 316, an optional notification component 317, at least one processor 318, and an optional battery 332 .

  The treatment delivery interface 302 (if included) may be one or more electrodes configured to provide treatment to the patient, such as, for example, one or more defibrillation electrodes 320, pacing electrodes 322 and / or TENS electrodes 324. Can be combined with Sensor interface 312 and treatment delivery interface 302 facilitate the exchange of data between sensor 212, sensing electrode 328, motion sensor 334, treatment delivery devices 320, 322 and 324, and medical device controller 300. Various connectivity and communication techniques may be implemented. In some examples, one or more of motion sensor / orientation sensor 334 includes one or more of motion sensor / orientation sensor 104 and / or motion sensor or orientation sensor 250. Also, in some instances, treatment delivery devices 320, 322 and 324 are included in treatment electrode 214.

  Data storage 304 includes one or more non-transitory computer readable media such as, for example, flash memory, solid state memory, magnetic memory, optical memory, cache memory, combinations thereof, and the like. Data storage 304 may be implemented for use in one or more of sensor placement 100, motion tracking unit 106 and wearable medical device 200 as well as the operation of medical device controller 300 itself in various examples. Configured to store instructions and data. Furthermore, data storage 304 includes motion information data store 330, as described in more detail below. Motion information data store 330 includes patient motion data collected by one or more motion sensors / orientation sensors 334 and transmitted from one or more motion sensors / orientation sensors 334. The motion information data store 330 includes information necessary for measurement and estimation of motion and at least one of motion primitives, motion modifiers, motion objects, motion grammars and motion sentences. Motion primitives, motion modifiers, motion objects, motion grammars are all referred to as "motion sentence features". In the case of a template based motion recognition system or a heuristic motion recognition system, this may include thresholds as well as other logic based elements and data structures. For deep learning networks, this may be a node factor. For the HMM approach, this may be in the form of state transition probabilities. The motion information data store 330 may also include data acquired during a baseline or training period, or data derived at least partially from data acquired during a baseline or training period. Predefined motion sentence features and motion sentences used to classify inbound motion data from one or more motion sensors / orientation sensors 334 while the motion classification component of the medical device controller is operating in the monitoring mode The corresponding motion data is further described below.

  In some examples, network interface 306 may facilitate information communication between medical device controller 300 and one or more other devices or entities via a communication network. For example, if the controller 300 is included in a portable medical device (such as the medical device 200), the network interface 306 may be within a hospital medical device such as the hospital medical device 1316 described further below with reference to FIG. May be configured to communicate with the medical device controller 300 included therein. In another example, the network interface 306 may be configured to communicate with a remote device that includes a processor, such as a programmable device described further below with reference to FIG. The caregiver may access information related to the patient at the remote device.

  In some examples, the optional user interface 308 is configured to cause operation of one or more physical interface devices such as input devices, output devices and combinations of input devices / output devices, etc. And the software stack being These user interface elements may display visual content, audio content and / or haptic content, such as content related to location specific processing. Thus, the user interface 308 may receive input or provide output, which allows the user to interact with the controller 300.

  The sensor interface 312 is coupled to one or more motion sensors / orientation sensors 334 and / or one or more of a sensing electrode / other sensor for receiving other patient data indicative of patient parameters. It may be combined in their combination. Once the data from the sensors is received by sensor interface 312, the source sensor is identified by processor 318 and the data is transmitted to the appropriate components in medical device controller 300. For example, if cardiac data is collected by the cardiac sensing electrode 212 and transmitted to the sensor interface 312, the sensor interface 312 transmits data to the processor 318, which in turn transmits the data to the cardiac event detector 326. Relay.

  In some implementations, processor 318 includes one or more processors. Each of the one or more processors is configured to execute a series of instructions that provide manipulated data and / or control the operation of other components of the controller 300. In some implementations, processor 318 makes certain logic-based decisions based on received input data when performing certain processes provided herein (eg, FIGS. 5 and 6). And may further provide one or more outputs. One or more outputs may be used to control or otherwise signal subsequent processing to be performed by processor 318 and / or other processors or circuitry communicatively coupled to processor 318. It can be used. Thus, processor 318 responds to a particular input stimulus in a particular manner and generates a corresponding output based on the input stimulus. In this sense, the structure of the processor 318 according to an example is defined by the flowcharts shown in FIG. 5 and FIG. In some illustrative cases, processor 318 goes through a series of logic transitions. At this logic transition, the states of various internal registers and / or other bit cells internal or external to processor 318 may be set to logic high or logic low. This particular sequence of logic transitions is determined by the state of the electrical input signal to processor 318, and when executing each software instruction of the software processing shown in FIGS. 5 and 6, the application specific structure is processor 318. It is assumed more effectively. Specifically, those instructions anticipate the various stimuli received and change the state of the associated memory accordingly. In this manner, processor 318 may generate and store useful output signals or may otherwise provide them. It will be appreciated that the processor 318 can process particular input signals and display particular output signals based on one or more logical operations performed during execution of each software instruction during processing execution. As mentioned herein, processor 318 is configured to perform a function when the software is stored in a data store coupled to processor 318, the software causing the execution of this function A series of various logic decisions are configured to be passed to processor 318. The various components that are described herein as being executable by processor 318 may be implemented in various forms of dedicated hardware, software or combinations thereof.

  In some instances, the controller 300 monitors the activity of the patient's heart, identifies cardiac events that the patient has faced based on the received cardiac signals, and, if necessary, one or more of the patient's body. Included is a cardiac event detector 326 for treating a patient by performing a treatment protocol that may result in multiple defibrillation shocks.

  In some examples, an optional motion recognition component or motion classification component 316 is configured to receive and process motion information obtained by one motion sensor / orientation sensor 334. In some examples, the motion recognition component or motion classification component 316 can be configured to operate in a baseline mode / training mode and / or a monitoring mode. FIG. 5 shows an example of a baseline processing / training process configured to be performed when motion recognition component or motion classification component 316 is operating in a baseline mode / training mode. FIG. 6 shows an example of a motion classification process configured to be performed when motion recognition component or motion classification component 316 is operating in a monitoring mode. The motion recognition component or motion classification component 316 is shown as optional. This is because, in some examples (eg, the example shown in FIG. 13), the motion recognition component or motion classification component 316 is implemented on a different programmable device (eg, server 1302). If the motion recognition component or motion classification component 316 is implemented by a device other than the medical device controller 300, the processor 318 may, for example, send inbound sensor data before transmitting sensor data to the remote motion classification component via the network interface 306. Data is stored in motion information data store 330 (eg, in sensor data structure 400). In some examples, the motion recognition component or motion classification component 316 uses the techniques described in US patent application Ser. No. 15 / 077,995, entitled System and Method for Determining Position Using a Medical Device. Are configured to determine the position of the medical device. The application is filed on March 23, 2016 and published as US Patent Application Publication No. US20160278652A1, which is hereby incorporated by reference in its entirety.

  In some instances, the optional notification component 317 may respond to the detected sequence of patient movements by alerting the alert, alert or other notification to justify the motion recognition component or motion classification component 316. It is configured to receive and process one or more notification requests from the motion recognition component or motion classification component 316 if one makes a determination. The notification component 317 may issue a notification via the network interface 306 and / or the user interface 308 when it executes according to its configuration. For example, notification component 317 may be a programmable device (eg, client device 1304 described below with reference to FIG. 13) and / or a medical device (eg, medical device 1316 described below with reference to FIG. 13) Alert a caregiver (eg, caregiver 1319 described below with reference to FIG. 13) to any of 1314, 1312, 1310, 1308) through the programmable device user interface One or more instructions may be sent. Alternatively or additionally, notification component 317 may indicate notification to patient 102 via user interface 308. These notifications may include visual, audio and tactile elements.

  In some implementations, in addition to, or as an alternative to, notification actions, notification component 317 may be configured to store the output of the motion recognition process or motion classification process in one or more databases. In such cases, this data may be grouped by type of motion data, when it was collected, and / or by annotations of this data based on analysis of the information. An authorized person, such as a caregiver or helper, may cause the medical device to generate one or more reports based on this data. For example, such reports may be ordered by data type and the date and time the data was collected.

  In various implementations, controller 300 implements an embedded operating system that provides file system and network support. In one example, controller 300 includes a relational database function, touch screen display driver, voice generation, Bluetooth® wireless network, iBeacon or near field communication methods such as active NFC beacons or passive, contactless RFID tags, etc. Including, but not limited to, Bluetooth® low energy (BLE) beacon technology, network security, firewall and software functions that provide data encryption services.

  Motion information is augmented by location beacons located at connection points in facilities such as hospitals, for example at the end of the entrance walkway, the stairway of the hospital, the cafeteria and entrance and in the middle aisle in the door of the patient room It can be done. Location beacons may be implemented with short range beacons such as iBeacon. The near field beacon communicates with the device worn by the patient, as well as providing absolute location position data that the device can use to calibrate its position, as well as provide an indication of its actual position. Send to a caregiver. By combining a motion sensor / orientation sensor with a beacon system, the number of beacons can be reduced, thus reducing the cost and complexity of implementation.

  FIG. 4 illustrates an example of a motion information data store 330, shown as a component of data storage 304 of medical device controller 300, for storing a plurality of motion parameters. Motion information data store 330 includes sensor data structure 400, sensor master structure 402, motion primitive master structure 404, motion primitive data structure 406, motion sequence master structure 408, and motion sequence structure 410. In some examples, data structures 402, 404, and 408 may be motion aware components or motions while operating in baseline mode / training mode (eg, during initial fitting of a medical device including medical device controller 300). It is populated by classification component 316. In some examples, data structures 400, 406 and 410 are populated by motion recognition component or motion classification component 316 when operating in a monitoring mode.

  The sensor master structure 402 stores a list of motion sensors / orientation sensors (eg, one or more motion sensors / orientation sensors 334 that acquire motion data and transmit to the processor 318 via the sensor interface 312). Examples of types of sensors represented in sensor master structure 402 include, but are not limited to, one or more of a one-dimensional accelerometer, a two-dimensional accelerometer, a three-dimensional accelerometer, a gyroscope and a magnetometer. As shown in FIG. 4, sensor master structure 402 includes at least two parameters including a sensor_ID field and a position field. The sensor_ID field is configured to store an identifier of each of the one or more motion sensors / orientation sensors 334. These identifiers are used throughout the data model shown in FIG. 4 to associate data with the identified sensors. The location field is configured to store data describing the anatomical location on the body of the patient 102 in which the sensor identified by the sensor_ID is located. Thus, the data stored in the location field associates the sensor with the location on the patient's 102 body (e.g., left forearm, right forearm, left shin, right thigh, crown, chest). This position data may then be used to identify motion primitives and, optionally, motion sentences performed by the patient 102, such as the motion information data store 330 and other elements of the medical device controller 300 (eg, motion). Used by recognition component or motion classification component 316). The sensor master structure 402 improves the computational efficiency of the medical device controller 300. This is because information that does not change frequently (e.g. sensor position) is not transmitted with each data item of motion data. In some examples, the sensor master structure 402 may include the actual sensor ID and anatomical location during initial fitting of a sensor arrangement (eg, sensor arrangement 100) or a portable medical device (eg, portable medical device 200). It is populated with

  The movement primitive master structure 404 stores baseline data describing various predetermined movement primitives that are recorded by the medical device controller 300 when being executed by the patient 102. As mentioned above, motion primitives are essentially "verbs" of motion sentences that include motion sentence features. The motion primitives are often simple enough that they can be executed by the patient 102 when commanded to collect corresponding patient-specific baseline data during the baseline / training period, or It is a predefined movement that is sufficiently general. Motion primitives may also be identified as more complex motion components, including other predefined motion sentence features in one or more motion sentences. Examples of motion sentences that are particularly useful in clinical practice (eg, hospitals or inpatient clinics) include, among other things, 1) "Get up from bed" = "Get up from bed" + "with both legs at the edge of the bed Go "+ put both feet on the floor" + "stand up" 2) "sit in bed" 3) "get up from bed in supine position" 4) "rest in bed from sit" 5) "left foot 6) "Put the right foot", 7) "open the door", 8) "turn while standing in the proper position", 9) "sleep asleep" and 10) "stand up from a chair or toilet Sit ". Any one or more of these exemplary motion primitives and motion sentences or other motion primitives, motion sentences and motion sentence features not explicitly identified herein may be motion sensor assemblies or orientation sensors An individual patient (such as patient 102) wearing assembly 100 or portable medical device 200 is recorded. In this way, the movements or features and their inherent movements of the individual patients and individual movement sensors / orientation sensors are reflected in the baseline data recorded during the baseline period / training period.

  Among other things, these motion primitives may be recorded and analyzed for each of the individual patients while the medical device controller 300 is operating in baseline mode / training mode, as described below in FIG. Thus, not only does the library of baseline data be generated, it results in an improvement in the accuracy of the motion recognition algorithm of that particular patient. Data from the motion sensor / orientation sensor is measured while the medical device controller 300 is operating in the monitoring mode and input to the motion recognition algorithm. Motion recognition algorithms classify motion such that the motion best corresponds to a predetermined motion primitive or motion sentence feature. Then, whether the output of motion recognition (MR) processing (eg, motion primitives, motion modifiers, motion objects, motion adjectives and motion grammar structures) pose a threat to the patient's health or safety, or If not, it can be analyzed to determine if it constitutes an early failure. Next, in response to the classification of the detection operation, a notification component (such as an alarm or notification to a healthcare provider or other caregiver) may be performed. An appropriate signal may be provided to one or more of the patient 102, the healthcare provider or components of the medical device system, any one of which may initiate an action. These actions may include protective measures if a series of motion primitives are expected that patient motion will result in dangerous results.

  As shown in FIG. 4, the motion primitive master structure 404 includes a primitive_name field, a primitive_ID field, a position field, a parameter_ID field, a time field, and a value field. The primitive_name field is configured to store data describing a human readable name for each motion primitive to be recorded for the patient 102. The primitive_ID field is configured to store an identifier of each motion primitive. These identifiers are used throughout the data model shown in FIG. 4 to associate other data with the identified motion primitives. The remaining four fields, ie, the position field, the parameter_ID field, the time field and the value field, are configured to store sensor data obtained during recording of each motion primitive. More specifically, each movement primitive is recorded at an anatomic position (specified in the position field) at a specific time (specified in the time field) of the parameter (identified in the parameter_ID field) It may consist of one or more sensor values (specified in the value field). The motion primitive master structure also includes information describing motion objects and motion modifiers.

  The motion sequence master structure 408 stores data describing motion sentences (both single motion sentences and a series of motion sentences). Examples of motion sentence names include "step right", "rise from bed", "turn (while standing)", "sit (from standing)" and other such examples. A series of motion sentences can explain the patient's more complex movements. An example of such a sequence is that the patient leaves the bed and goes to the bathroom, the patient walks the corridor away from his room, the patient (the steady walk is gradually decreasing over time Performing a 6 minute walk test) which may be used as patient specific baseline data acquired during a baseline period / training period to identify. As shown in FIG. 4, motion sequence master 408 includes a Sequence_Name field, a Sequence_ID field, a Primitive_ID field, and an Action field. The Sequence_Name field is configured to store data describing a human readable name for each motion sentence associated with the patient 102 while the patient is being monitored by a medical device including the medical device controller 300 Be done. Examples of series of motion sentence names include "go to bathroom", "perform a six minute walk test", "go to nurse station" and other such examples. Thus, each sequence consists of two or more motion sentences that together describe the entire sequence of actions.

  The Sequence_ID field is configured to store an identifier of each motion sequence. These identifiers are used throughout the data model shown in FIG. 4 to associate other data with the identified motion sequence. The Primitive_ID field is configured to store one or more Primitive_IDs that together make up the motion sequence identified in the Sequence_ID field. The action field is configured to store data identifying one or more actions to be triggered in response to the identification of the motion sequence identified by the sequence_ID. Actions identifiable through this field in some instances include the execution of a notification component. The implementation is an automation designed to handle the result of the detected motion. For example, as described further below with reference to FIG. 13, the notification component may be executable by the processor to alert the device associated with the user or located at a particular location Configured as.

  The action field may also be configured to include various conditions that affect which of one or more actions are triggered when. For example, the action field may include the patient's age, the current time of day (including designated day and night times and times) associated with the movement, the operation mode of the movement classification component and / or the action to be performed. . For example, an alert is sent to the pager of a specialized nurse who is traveling for a 90-year-old patient during the daytime when the patient is out of bed. During night zones with less staff, similar alerts may be triggered and sent to the nurse station. Alternatively, for a 20-year-old patient, no alert is triggered regardless of the time the patient is out of bed.

  The sensor data structure 400 stores various data describing patient motion acquired by each motion sensor or orientation sensor while the medical device controller 300 is executing the monitoring mode. As shown in FIG. 4, the sensor data structure includes a sensor_ID field, a parameter_ID field, a time field and a value field. The sensor_ID field is configured to store an identifier of a motion sensor or orientation sensor, as defined in sensor master structure 402. The parameter_ID field is configured to store data specifying identification information of the parameter measured by the data value acquired by the sensor identified in the sensor_ID field. For example, force data in the x-axis, y-axis and z-axis directions each have different parameter IDs that specify the direction of the force. Similarly, the orientation (eg, tilt, rotation) data transmitted by the gyroscope has one or more parameter IDs used to identify the measurement corresponding to that data. The time field is configured to store data indicating when each measured value was acquired by one or more motion sensors / orientation sensors and / or the duration that motion is taking place. The time fields of sensor data structure 400 are used as an index to arrange the sensor data collected for patient 102 in an orderly manner. Thereby, to determine that a motion sentence or a plurality of motion sentences can be placed in a motion sequence, and also to measure the duration of the motion (e.g. from the start of the motion to the end) , Multiple movements can be analyzed. The value field is configured to store measurements of value parameters obtained by the motion sensor / orientation sensor. These values may be, for example, scalar or vector values measured by the motion or orientation sensor as describing the motion of the patient 102. It is understood that scalar values may be associated with one or more vector components when analyzed with one or more of sensor_ID, parameter_ID and other data structures and / or elements of motion information data store 330. You see.

  Motion primitive data structure 406 stores data describing motion primitives and other motion sentence features shown in the sensor data stored in sensor data structure 404. As shown, motion primitive data structure 406 includes a primitive_ID field, a position field, a parameter_ID field, a time field, and a value field. In some instances, the data stored in these fields combines the data stored in sensor data structure 400 with the data stored in sensor master structure 402 and matches the associated position, parameter_ID, position It is determined by identifying the Primitive_ID from the Motion Primitive Master Structure 404, having a time and value, a parameter_ID, a time and a value in the combined data. The degree of confidence needed to find a match may vary between examples.

  The motion sequence structure 410 stores data describing a motion sentence and a series of motion sentences shown in the motion primitive stored in the motion primitive data structure 406. As shown, motion sequence structure 410 includes a sequence_ID field and a primitive_ID field. In some examples, the data stored in these fields combines the data stored in motion primitive data structure 406 with the data stored in motion sequence master structure 408 and with the primitive_ID in the combined data. It is determined by identifying the sequence_ID from the motion sequence master structure 408 that has a matching associated primitive_ID. The degree of confidence needed to find a match may vary between examples. It will be appreciated that in some examples, motion primitive data structure 406 and motion sequence data structure 410 may be implemented as a view to sensor data structure 400 rather than a data store that includes copies of data.

  A state space may be defined if there are two or more relevant patient motion states. These patient motion states may correspond to one or more motion primitives described herein. Exemplary patient motion states are shown below for illustrative purposes only. Some deviations of the stated measurements can be expected in various cases.

  Patient Sitting: Inactive-The patient's torso is inclined at an angle greater than 15 degrees, but 1 foot / second (or for example 0.25, 0.5, 2, etc.) for a continuous period of 5 seconds There is a moving speed below the other appropriate threshold such as 4, 8 or 16 feet per second. In other examples, the sitting patient: inactive definition may be based on other movement measurements, such as root-mean-square (RMS) movement, acceleration, etc., or a multi-dimensional combination thereof. The situation in which the patient is sitting may also have a sub-state of "active" with a movement speed greater than that defined for the "inactive" state.

  Patient is asleep: inactive-patient is inclined less than 15 degrees but less than 0.25 feet per second (or other threshold, range, parameter measured as above) lasting 10 minutes There is a moving speed of

  Patient is asleep: Active-patient is inclined less than 15 degrees, but at 0.25 feet per second (or other threshold, range measured as above, for the last 10 minutes) Parameter) There is a moving speed of over.

  Patient is out of bed-this condition is that the patient is sleeping: inactive or the patient is sleeping: either one is ahead. The patient is inclined more than 15 degrees. When the patient is out of the bed, it is a process of “raising up”, which is a process in which the body tilts upward to a vertical position, and the motion sensor / orientation sensor 104 located on the patient's foot or ankle is 6 inches downward. It can be further subdivided into the state of putting both feet down, which is ultra-lowering. In some instances, “the patient puts his foot on the floor” is a motion or orientation sensor located either on the patient's thigh or torso, aligned vertically less than 0.5 feet to the foot's sensor It is shown when it happens.

  Patient is asleep: Turnover-Patient's turnover sub-state: "sleeping" means that the angular rotational movement of the torso is a rotation angle of 20 (while still remaining at an angle of less than 15 degrees to the horizontal) It is defined when the degree is exceeded. The patient is asleep: If a turn over is detected, the device has a physiological monitoring sensor (eg ECG, SpO2, etc.) currently under the patient and is compressed between the patient and the mattress It can be judged that there is a possibility. In response, different sets of sensors having positions that facilitate more accurate monitoring of the patient's physiological metrics may be activated (eg, they are not compressed between the patient and the bed) That is the reason). For example, the ZOLL LifeVest® Wearable Defibrillator includes two sets of two ECG leads, an ECG electrode pair connected to a pair of front and rear (FB) leads and left and right (SS) leads. If it is determined that the patient is lying on one side down, the processor will switch to monitoring SS leads instead of FB leads, or alternatively, greater compression on FB leads And some additional processing may be performed, such as signal filtering in anticipation that motion artifacts will occur.

  The patient is standing up-similar to the patient being out of bed, but not in the patient's sleeping state, but in front of the patient's sitting.

  The patient is standing-the patient's torso is less than 5 degrees to the vertical, but the rate of movement of the horizontal movement is 1 ft / s (or other suitable for the duration of the last 5 consecutive seconds) (E.g., 0.25, 0.5, 2, 4, 8 or 16 feet per second).

  The patient is falling-the previous condition is that the patient is standing or sitting and in these conditions the angular velocity to the horizontal is more than 10 degrees per second.

  The patient is walking-the patient's torso is less than 5 degrees to the vertical, but the rate of movement of the horizontal movement is 1 ft / s (or other appropriate) in the last 5 seconds A threshold, for example 0.25, 0.5, 2, 4, 8 or 16 feet / second, is exceeded. Techniques known to those skilled in the art may be used, such as Zero Speed Update (ZUPT) or Zero Speed Detection (ZVD). This causes the state machine to determine when a particular motion sensor / orientation sensor 104 on the patient has stopped moving for a sufficient amount of time to perform the calibration. The intervals may be regular or irregular. For example, the calibration may be based on known motion states such as "patient is walking". In this case, the calibration may be based on motion estimation relying primarily on the motion sensor / orientation sensor 104 located on the patient's foot or ankle, with zero velocity if the patient places a particular foot on the ground at each step. What is described by Skog et al. (Herein “Zero Speed” by I. Skog, P. Handel, J.-O. Nilsson and J. Rantakokko, which is incorporated herein by reference in its entirety) Detection—algorithm evaluation ”(November 2010, IEEE Trans. Biomed. Eng., Vol. 57, No. 11, pp. 2657, pp. 2666)) and the like. According to Skog et al. "Even if the motion information provided by the footed inertial sensor is used for different purposes in gait analysis and pedestrian navigation, high quality motion information is needed in both applications."

  Further, Skog et al. "By utilizing a low cost sensor for Inertial Navigation System (INS), the position error is proportional to the cube of the operating time. Free Inertial Navigation can only be realized for periods within a few seconds However, third-order error growth can be reduced by imposing constraints on the navigation solution using information about the dynamics of the system The type of information generally used for this purpose is the knowledge of the time when the system is in a stationary phase, ie the time when the system has a constant position and attitude.This information to limit error growth. The use of zero is referred to as the use of zero speed updates, as the foot periodically returns to rest during normal walking. B speed update is described as well suited to limit the error growth of the foot with INS. ".

  Further, Skog et al. "A 'soft' zero speed update is commonly used in footed inertial sensor based pedestrian navigation systems, where the cumulative travel of the foot over several steps is targeted. That is, to provide an estimate of the accumulated error from the last zero speed update, knowledge of when the system has zero speed is a measure of how position errors, speed errors and attitude errors are over time. It is used with a model of what happens. An estimate of the accumulated error is then fed back to correct the navigation solution and to calibrate the navigation algorithm. It is explained that.

  Skog et al. Described “hard” zero velocity updates as “hard” zero velocity in gait analysis in which foot movements during individual walking cycles are targeted, not cumulative movement distance of the foot over several cycles Updates are generally used. When the system imposes a zero velocity update, the update is hard in that position, velocity and yaw are reset to zero and roll and pitch are initialized directly from the accelerometer's measurement of gravitational acceleration. It is explained that.

  While the device appears to be inactive from the user's point of view, but is actually performing various self-tests that do not require intervention by the user or patient (eg, (“The patient is sleeping: The calibration of the motion sensor / orientation sensor 104 may also be performed during either an inactive "or an inactive sub-state (such as when the patient is at rest: inactive). During these periods of inactivity, calibration intervals may be performed as often as every millisecond, but may be performed as often as once every minute or even once every three hours. The calibration may last for several periods, such as one second, one minute, one hour, etc., during which time dependent factors such as offset drift may be estimated.

  During these calibrations, motion features such as RMS noise, offset and drift may be estimated. The updated estimate of RMS noise may be used to determine a threshold for detection of zero speed conditions. The updated estimates of offset and offset drift can be used to reduce the final position accuracy to approximately 1% of the total travel distance (eg, for every 30 feet of travel, the error is approximately 3.5 Will be in inches).

  Exemplary Motion Classification Process FIG. 5 is a flow illustrating a patient-specific baseline processing / training process 500 for collecting baseline data corresponding to motion primitives, motion sentence features, motion sentences, and a particular series of motion sentences. FIG. The patient specific baseline processing / training processing 500 may be performed by a medical device that includes the medical device controller 300 or another programmable device that includes a motion recognition component or a motion classification component 316. The motion recognition component or motion classification component 316 may operate in a patient specific baseline mode / training mode. The patient-specific baseline processing / training process 500 places 502 one or more motion sensors / orientation sensors (eg, one or more motion sensors / orientation sensors 334) on the patient (eg, patient 102); Beginning by recording the position of each of the one or more sensors in the sensor master structure 402. This recording may be performed by default or by positively assigning a position to the sensor via the user interface. If the predefined position of one or more sensors is not corrected via the user interface, then the position data stored in sensor master structure 402 requires that a particular sensor be located at patient 102 at a particular position. obtain. In some instances, these sensors include visual indicia of the appropriate anatomical location for the sensor.

  The patient-specific baseline processing / training processing continues to identifying movement primitives to be performed by the patient such that corresponding baseline data can be collected and analyzed (504). The identified (504) motion primitives may be any of the examples described above or other motion primitives used to identify and classify patient motion. Next, the patient is prompted to execute motion primitives (506). This prompt (506) may be provided directly to the patient 102 by the medical device (or similar device in communication with the motion information data store 330) or by the caregiver collecting baseline data using the medical device . When the patient 102 executes a motion primitive, a motion sensor or orientation sensor (eg, one or more motion sensors / orientation sensors 334) attached to or attached to the patient 102 may be sensor data corresponding to the motion primitive Record (508). Next, sensor data is stored (510), for example, in the motion primitive master structure 404, along with the identified (502) and executed (504) motion primitives. This establishes baseline data corresponding to identified (502) and executed (504) motion primitives, including motion objects, motion modifiers, motion sentences and sequences of motion sentences.

  Next, baseline processing / training processing 500 determines whether baseline data for all motion primitives requested or recommended to the patient has been recorded (510). A list of all movement primitives required or recommended to be recorded: age, weight, height, health risk (loss of orientation, fatigue, memory loss, etc.), type of illness, type of surgery, or The patient 102 may be identified based on various health factors, such as future or previously performed treatments and others. Still other processing may be required as part of a standard process to monitor patient movement, to determine deviation from standard movement (e.g. to determine cough to determine seizures) To determine the tremor). If all of the required or recommended motion primitives have been recorded for the patient, method 500 ends. If some motion primitives remain to be recorded 510, the patient-specific baseline processing / training process 500 returns to the identification 502 of the next motion primitive to be recorded. Next, the aforementioned aspects of the patient-specific baseline processing / training process 500 are repeated until all motion primitives are recorded and stored (510).

  The patient-specific baseline processing / training process 500 described above has been described in terms of single motion primitives, motion objects and motion modifiers (eg, sit, step left leg, step right leg), but the baseline It will be appreciated that data may be recorded and stored 510 for a series of motion sentences. For example, baseline data may be recorded 510 for a series of actions, such as getting up from the bed and subsequently moving the legs to the side of the bed, then getting up and then taking off. In another example, baseline data is recorded for a sequence of walking to the patient's room door, opening the room door, and walking out of the room. In yet another example, a series of motion primitives and motion sentences and a series of motion sentences are recorded (510) while the patient is performing a six minute test. Other sequences will be understood in light of the description herein.

  FIG. 6 is a flow diagram illustrating one possible monitoring process 600 for measuring the output of the motion sensor and classifying the motion or orientation as corresponding to a motion primitive, motion sentence or sequence of motion sentences.

  Monitoring process 600 may be performed by a medical device that includes medical device controller 300 or another programmable device that includes motion recognition component or motion classification component 316. In some implementations, the monitoring process 600 may be an external programmable device, such as a telephone, a personal digital assistant, a tablet, or other electronic that is remote from the medical device but may also be carried by a patient wearing the medical device. It can be configured to occur on the device. In some instances, such external devices may not be carried by the patient, but may also be carried by the caregiver, the patient's agent (eg, a lover) or other individuals who may accompany the patient as they move about. . In one scenario, a nurse in a hospital environment can carry such a device while assisting a patient who can move in a hospital building. For example, such phones, personal digital assistants, tablets or other electronic devices are operatively connected to the medical device (eg, via a wired or wireless connection) to exchange information regarding the monitoring process 600 with the medical device It can.

  In some implementations, the monitoring process 600 may be configured to occur at a remote server (eg, a medical device facility or a caregiver facility). For example, such a remote server wirelessly connects with the medical device (eg, via a BlueTooth® activated base station or WiFi connection) and exchanges information regarding the monitoring process 600 with the medical device It can.

  Motion recognition component or motion classification component 316 may operate in a monitoring mode. The monitoring process 600 begins by acquiring and / or measuring (602) sensor data describing that the patient 102 is moving or stationary, using a medical device. Sensor data is communicated to the medical device (or similar device that is part of a distributed computing network) and recorded in data store 330 (eg, sensor data structure 400). Next, motion classification or motion recognition 604 may be any of a variety of machine learning techniques including artificial neural networks, deep learning networks, HMMs, feature vector analysis and distance function analysis, cross correlation, threshold analysis or the details described above. This may be performed (e.g., by motion recognition component or motion classification component 316) on sensor data using other known techniques. If a statistical or probabilistic approach is taken for motion classification processing or motion recognition processing 604, the confidence level of the classification is determined as part of the operation of machine learning or other classification techniques. If the determined confidence level 608 is below a threshold that sets the minimum match level required to find a match, the sensor data corresponding to the detected movement is flagged for verification by the caregiver Is set (606). The monitoring process 600 ends after the flag setting (606) of sensor data.

  In the case of the statistical recognition approach, if the determined confidence level 608 is greater than the confidence threshold, then the classified (604) motion sentence features represent the current motion sentence (eg, stored motion primitive data The structure 406) is added (610). In the template matching approach to motion recognition, motion sentences and sequences of motion sentences obtained during baseline processing / training processing are stored as patient-specific baseline data (eg, as stored in motion sequence master 408) It can be done. The measured motion, detected motion sentences features and motion sentences can then be matched against general motion sequence master data as well as patient specific baseline data to find the best match with motion sequence 612 . For the sake of illustration, the classified (604) motion sentence features are the motion primitive "rise" and the motion modifier "rise", and the previous orientation (past history) is "lying" and preceded by If "sleep", classification step 612 determines that the complete sentence "is up from the bed" based on both the universal motion database, the patient specific baseline data and the universal grammar rules. If the motion sequence is not recognized 614, the motion sequence is flagged 618 for verification by the caregiver and the monitoring process 600 ends. However, if one or more defined motion sentences or motion sentence sequences have been identified (ie, not undefined (614)), the monitoring process 600 may perform a series of movements by the patient (eg, It is determined whether the motion sequence master 408 has data describing the action stored therein, or whether it is a motion sentence or a series of motion sentences that expose the patient to an initial or immediate risk. If so, the monitoring process 600 performs a notification action (eg, a notification component, eg, sending a notification request to the notification component 317). Examples of notification actions that may be requested to perform 620 via the notification component include alerting the patient, a healthcare professional, or both, of a fraudulent or high risk motion. If the sequence of motion primitives is not feasible, monitoring process 600 performs operation 602. At act 616, the monitoring process determines if the motion sequence is complete (616), for example by using stored grammar rules. If it is determined that the motion sequence is complete (616), a new motion sequence is initiated (eg, a new record is added to the motion sequence structure 410) and the monitoring process 600 performs operation 602.

  Returning to operation 602, the monitoring process 600 continues by waiting for another patient's movement to be detected via the acquisition of more sensor data (602). In some instances, the actions of flag setting 606 and 618 described above may alert the caregiver associated device to an emergency alert to alert the caregiver that an unidentifiable motion has been performed by the patient 102. It will be understood that it may include emitting.

  In some instances, whether the motion sequence is feasible or not is based at least in part on contextual factors. For example, in various instances of action 616, motion recognition component or motion classification component 316 may further provide prescribed medical instructions (eg, in consecutive beds) regarding the time of day, the patient's age, the patient's medical history, and the patient's movement. Assessing whether there has been a break, a violation of not walking for more than 10 minutes at a time, and / or a motion recognition component or motion classification component 316 operation mode. In some examples, motion recognition component or motion classification component 316 can be configured between daytime monitoring mode and nighttime monitoring mode. In these examples, some motion sequences may be feasible in the nighttime monitoring mode, but not in the daytime monitoring mode or vice versa. For example, while the motion sequence of the patient going to the bathroom may not be feasible while the motion recognition component or motion classification component 316 is operating in daytime monitoring mode, the same motion sequence may be motion recognition component or motion classification It may be feasible if component 316 is operating in the nighttime monitoring mode. In some examples, the motion recognition component or motion classification component 316 can be used between a standard monitoring mode and a delivery mode of fluids or drugs into a vein, a monitoring mode associated with a medical procedure such as dialysis or other medical procedures. It can be set. For example, the motion sequence of a standing patient may not be feasible during the standard monitoring mode, but dialysis or some other treatment (eg, sedatives) where the patient is prescribed rest in bed The same motion sequence may be feasible if the patient is receiving a dose).

  Not all motion primitives need to be recorded for all individual patients. For example, when there is an unsupported fall to the ground, most of the parts are (a) unique among other movement primitives and (b) general sensor data is generated for most patients . For this reason, existing generics that may be in the form of models generated from multiple patient and motion databases, heuristic rules, or recognized typologies from data previously collected and entered into HMMs, deep learning networks The data may in some cases be used for recognition of some motion sentence features, motion sentences or motion sentence sequences. For example, as a result of the specificity and commonality of sensor data describing patient falls, the classification process used may be more computationally specific and efficient.

  FIG. 7 is an example of existing generic data that may be used to identify a series of motion primitives presented by the patient without the patient having to execute the motion primitives during baseline processing / training processing. FIG. 7 is a graph of the force acting on the patient as a function of time during a fall to the ground in an unsupported state. The forces shown are directed in the x, y and z directions in a Cartesian coordinate system. The sum of the absolute values of these forces is also shown. Although the exemplary data shown in FIG. 7 was collected from a three-dimensional accelerometer, an alternative representation of the physical phenomenon of falling to the ground in the unsupported state of the patient is used and the sensor and sensor orientation It will be understood that it is possible depending on. The positive and negative signs of the data shown in FIG. 7 were a function of the orientation of the 3D accelerometer. For the purpose of illustration, only the magnitude as a function of time is stated.

  As shown, at time point (t) = 0 seconds (s), the patient was standing upright. While enduring gravity, the patient (or more specifically, the patient's attached accelerometer) received a longitudinal force equal to 1 G (negative y-direction due to the orientation of the accelerometer on the patient). This 1 G force continued until approximately t = 0.35 s when the patient's unsupported fall to the ground (also known as "free fall") began. During free fall from t = 0.35s to approximately t = 0.48s, the net force acting on the patient decreased. Because the patient is no longer able to bear against gravity being pulled up by standing up. This reduction continued until the patient touched the ground. The various forces acting on the patient were approximately zero just before the patient hit the ground.

  FIG. 8 shows data similar to that shown in FIG. 7, but on a longer time scale. FIG. 8 further illustrates the forces experienced by the patient during and after a collision with the ground. As shown in FIG. 8, patient movement began at steady state with a net force acting on the patient of approximately 1 G. When the patient fell freely towards the ground, the force decreased for the same reasons as described above. However, as shown in the area labeled "collision", once the patient was in contact with the ground, the force of the collision was over 10G. In the example shown, when the patient bounces off the ground, and when the force of the collision echoes throughout the patient's body, a continuous wave of great force is generated. The force of the collision eventually dissipated as shown in the area labeled "Stable" after the collision.

As with other motion primitives and their motion sequences, classification of motion primitives and motion sequences may be realized by using sensor data and comparisons with various thresholds. Should the motion sequence be classified using this threshold comparison technique, the medical device controller 300 responds to the data across one or more of these thresholds (patient, monitoring system and / or caregiver ), Alerts or alarms may be triggered via a notification component (eg, notification component 317). In the case of free fall, a series of motion primitives corresponding to forces are stored as existing generic data, and the threshold force is set to detect a patient in free fall. These thresholds may include forces that increase or decrease continuously. When each threshold is exceeded, a new alert or alarm is triggered by the motion classification component via the execution of the notification component. Examples of thresholds for falls may include 0.8 G ₀ , 0.5 G ₀ and any value therebetween that indicates that the patient is moving towards the ground. In another example, the threshold indicating the stability after the collision may be any value between 0.95 g ₀ and _1.05-0. Also, as mentioned above, confidence metrics may be determined from motion sensor data or orientation sensor data and used as an additional threshold for engaging the notification component. In an example, a mobile portable device such as the exemplary medical device 200 may request the patient to confirm a fall, notify an onlooker (eg, through an alarm), and / or medically indicate that the patient has fallen. It can also include a system that can notify the provider.

  FIG. 9 is a schematic diagram of a motion sequence 900 showing a fall of the patient. As shown, the first motion primitive of the sequence 900 includes a free fall motion primitive 902, a collision motion primitive 904, and a stabilized motion primitive 906. In conjunction with the monitoring process 600, each of these motion primitives can be classified at operation 604 using threshold comparison techniques, which eliminates the need for patient-specific baseline data. Further, monitoring process 600 may determine that any motion sequence including free fall motion primitives is feasible at operation 616, and in response, takes action at operation 620. This action may include the execution of a notification component configured to alert or warn the patient and / or the caregiver about the fall. In addition, monitoring process 600 may take additional actions when motion primitives 904 and 906 are added to the motion sequence.

  In some instances, movement primitives used to determine whether a patient has fallen may include data describing orientation. For example, if the patient's orientation suddenly changes in a short period of time from substantially longitudinal to substantially transverse, the examples disclosed herein identify this series of patient motion as a fall; Actions can be taken accordingly.

  Similar to the patient fall example described above, the existing generic motion sentence features, motion sentences and series of motion sentences are applied to the patient from simulated data or data generated by one or more other patients. It may involve gaining, directionless movement, noise and fainting. Similarly, as shown in FIGS. 10, 11 and 12 and described in more detail below, existing general purpose motion sentence features etc may be simulated or generated for tremor, cough, seizure occurrence.

  FIG. 10 is a graph of the vector sum of patient motion acquired during the onset of patient tremor with three different motion sensors / orientation sensors located at three different locations on the patient. In the example shown, tremor was detected as high frequency (e.g., approximately 50 Hz to approximately 100 Hz or approximately 100 Hz to approximately 1000 Hz) vibrations. The main frequency may shift over time, but not in a short time. In some instances, high frequency vibrations are detected in some or all of a plurality of axes. In these multiple axes, the motion sensor or orientation sensor is configured to detect motion (eg, a one, two or three axis accelerometer), such that the motion may be detected (eg, motion sensor or Positioned so that the movement of the orientation sensor is not restricted between the patient and the bed). In some instances, the magnitude at different locations or different axes may be different, even though the net force on the patient is approximately zero.

  FIG. 11 shows the patient acquired during the onset of a patient's cough with three different motion sensors / orientation sensors located at three different positions on the patient (two on the torso and one on the knee) Is a graph of the vector sum of the motion of As shown, the pattern of force from coughing can be shown as acceleration of large forces in some directions. The acceleration of these large forces is temporary. That is, as a function of time, the force detected by the motion or orientation sensor on the patient's torso lasts for several seconds. As each cough accompanies compression of the patient's diaphragm, motion sensors located on the torso close to the diaphragm are more likely to detect the force caused by the cough.

  FIG. 12 is a graph of the vector sum of patient motion acquired during the onset of a patient's seizure with three different motion sensors / orientation sensors located at three different locations on the patient. In the example shown, the vector sum of seizures shown in FIG. 12 is such that periodic fluctuations in force are detected by the motion sensor / orientation sensor. Unlike the periodic frequency (shown in FIG. 11) detected from the cough, the periodic frequency of the exemplary seizure shown in FIG. 12 increases in amplitude as a function of time and has a time scale longer than cough. It happens over The movements resulting from the various types of seizures may differ between patients and among the seizure causes. For example, movement primitives detected during a seizure of fever may be different than those detected during a large seizure. It will be appreciated that it is unlikely that different types of seizures may utilize different existing generic data, that is, that it is unlikely that patient specific baseline data can be obtained for many of these types of situations.

  Net forces as a function of time are identified for each of the motion or orientation sensors in some instances, regardless of whether tremor, cough, seizures, snoring or loss of orientation are analyzed. It is useful to analyze the vector sum from each of the motion or orientation sensors to In some cases, it may be stationary to further identify the patient's movement that occurs at the appropriate position (eg, shaking while lying on a bed) or patient movement that occurs during walking of the patient. A reference motion sensor or reference orientation sensor may be used (eg, attached to a near stationary surface such as a bed or a relatively stationary surface such as a trolley with a carriage or an ambulance). Alternatively, or simultaneously, the force vectors not summed are analyzed to determine if the patient's movement is asymmetric (eg, there is a seizure only on the left side of the patient or no movement on the left side of the patient) It can be done.

  Each of the processes disclosed herein represent a particular set of acts in a particular example. The operations included in these processes may be performed by or using one or more programmable devices specifically configured as described herein. Some operations are optional and therefore may be omitted according to one or more examples. Additionally, the order of operations may be changed or other operations may be added without departing from the scope of the systems and methods described herein. Further, as mentioned above, in at least one example, the operations are performed on a particular specially configured machine, ie, a medical device configured in accordance with the examples disclosed herein.

  Exemplary Hospital Environment FIG. 13 is a schematic diagram of an example 1300 of a distributed computer system within a hospital environment in accordance with an example of the present disclosure. Example 1300 includes programmable devices 1302 and 1304, network 1306, and various medical devices 101, 1308, 1310, 1312, 1314, and 1316. For example, programmable devices in the network 1306 are secure and authorized to be used by a remote server 1302 and a caregiver or other authorized personnel to access the network 1306, as described in further detail below. And one or more client devices 1304, such as personal digital assistants or Internet enabled smart phones. In the example according to FIG. 13, the medical device 101 is associated with the patient 102 and includes at least one of the sensor arrangement 100 and the portable medical device 200. Example 1300 shows various locations within the hospital environment 1318, 1320 and 1322 where the patient 102 and the medical device 101 move with time.

  The server 1302 includes any computing device including one or more processors and is configured to communicate with one or more remote computing devices through the network 1306. The server 1302 also includes the motion recognition component or motion classification component 316 described above. In some examples, server 1302 maintains a local motion information data store that includes motion information stored in motion information data store 330.

  Client device 1304 may include various computing devices that may be arranged to communicate with network 1306. In some examples, client device 1304 is a computing device that can not only receive user input, but can also send and / or receive data via network 1306. For example, client device 1304 may be carried by a patient. For example, client device 1304 may be carried by a caregiver such as a nurse or a physician. In one embodiment, client device 1304 is a conventional computer system such as a desktop computer or laptop computer. In another embodiment, the client device 1304 is a device having computer functionality, such as a personal digital assistant (PDA), a cell phone, a tablet computer, a smartphone or similar device. In one example, client device 1304 can allow a user (e.g., caregiver 1319) to interact with network 1306 and receive data and analysis results of various patient motion primitives and / or motion sequences. Run the application to make it a dedicated computing machine. In at least one example, client device 1304 includes a notification component 317 located at the caregiver station and configured to receive a notification request from a motion classification component (eg, a motion recognition component or motion classification component 316). The client device 1304 may perform one or more client device actions in response to receiving the request, including, for example, presenting one or more notifications to a caregiver or patient.

  In some implementations, remote server 1302 may be configured to store the output of motion recognition processing or motion classification processing in one or more databases. In such cases, data may be grouped by type of motion data, date and time the data was collected, and / or annotations of the data based on analysis of the information. An authorized person, such as a caregiver or helper, may cause server 1302 (eg, via client device 1304) to generate one or more reports based on the data. For example, such reports may be ordered by data type, date and time the data was collected.

  Network 1306 may comprise any combination of local area network and / or wide area network using both wired and wireless communication systems. In one embodiment, network 1306 uses standard communication techniques and / or standard communication protocols. Thus, network 1306 is a technology such as Ethernet, IEEE 802.11, World Wide Interoperability for Microwave Access (WiMAX), 3G, 4G, CDMA, Digital Subscriber Line (DSL), etc. It may include the link used. Similarly, the network protocols used on network 1306 include Multi-Protocol Label Switching (MPLS), Transmission Control Protocol / Internet Protocol (TCP / IP), User Datagram Protocol (UDP), Hypertext Transport Protocol (HTTP), It may include Simple Mail Transfer Protocol (SMTP) and File Transfer Protocol (FTP). Data exchanged on network 1306 may be represented using techniques and / or formats including Hyper Text Markup Language (HTML) or Extensible Markup Language (XML). In addition, all or some of the links may be encrypted using conventional encryption techniques such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Internet Protocol Security (IPsec).

  In some instances, medical devices 1308, 1310, 1312, 1314 and 1316, which may communicate with network 1306 and may communicate directly with network 1306 and with other devices that may communicate with network 1306, are any medical devices or medical devices. Can be an assembly of Examples include, among other things, cardiac defibrillators, "emergency carts" stored on cardiac defibrillators, and various other devices used to resuscitate patients facing cardiac arrest. , Medicines and equipment, a wheelchair, and a trolley-mounted stretcher or mobile hospital bed. These devices, as well as others not expressly mentioned herein, are configured to communicate with the network 1306 and the motion recognition or classification component 316 within the server 1302 as described below. obtain.

  At a first exemplary location 1318, a patient 102 wearing a medical device 101 is exiting a hospital bed 1328 equipped with a unique motion sensor / orientation sensor 104. As noted above, relative movement between the medical device 101 worn by the patient 102 and the motion sensor / orientation sensor 104 attached to the stationary hospital bed 1328 is detected. The detected motion data is transmitted over network 1306 to server 1302 using the methods and systems described above and is classified by motion recognition component or motion classification component 316. In one example, the motion sensor / orientation sensor of the medical device 101 determines the number of steps taken by the patient using a series of motion primitives (eg, patient-specific baseline data recorded during a six minute walk test). This data may be sent to the motion recognition component or motion classification component 316 used for. The distance between the patient 102 and the hospital bed 1328 is estimated using the determined number of steps. Actions such as notifying hospital personnel of the patient's movement are performed based on this estimate and any pre-set threshold values for one or more allowable distances between the patient 102 and the hospital bed 1328. obtain. Similarly, an onlooker can also be notified to help the patient 102 return to the bed. In some examples, the motion sensor / orientation sensor 104 attached to the hospital bed may be one or more directly connected to the motion / orientation sensor 104 (or similar inertial measurement unit (IMU)) of the hospital bed 1328. Communicate with the processor. The processor and / or motion sensor / orientation sensor may also include a network interface. This configuration may improve response time and computational efficiency when using motion sensor / orientation sensor 104 of hospital bed 1328 to monitor patient motion.

  Examples of patient movements that may be monitored include gait as described above, but also include lesser patient movements. It may be helpful for these movements to have a stationary reference point, such as a movement sensor / orientation sensor 104 attached to the hospital bed 1328. These patient movements include tremors, coughs, seizures, loudness, fainting or other directionless movements. Having a stationary reference point is one example that may be used to detect these movements, but is not required. Other examples are walking, sitting in bed, standard to detect patient movements more accurately and classify them as tremors, coughing, seizures, roaring, fainting or other directionless movements. It will be appreciated that it may be dependent on baseline data recorded, measured and analyzed during a baseline or training period involving the patient in various activities such as breathing, sleep and other similar activities. In some cases, the existing generic data used for these are either generated from a model or generic data (such as in the examples of FIGS. 7 and 8) and specific to the patient 102. is not. In yet another example, the processor associated with the hospital bed motion / orientation sensor 104 or the processor associated with the server 1302 measures the magnitude of the patient's motion as the first threshold and the frequency of the patient's motion. The movement of the patient is at least partially classified by comparison with a threshold of two.

  In another example, depending on the classification of the patient's motion, the motion recognition component or motion classification component 316 identifies a medical device, such as the medical device 1316, associated with the patient's condition and in proximity to the patient, and notifies component 317 Configured to instruct the medical device to emit an alert via For example, if the classified motion is that of the patient 102 who has left the bed and is then in cardiac arrest (detected by the medical device 101), then the motion recognition component or motion classification component 316 is classified In response to movement and detection of cardiac arrest, an emergency cart proximate to the patient, such as the medical device 1316, is identified. If so equipped, the medical device 1316 (identifies the coordinates in the hospital to the client device 1304 associated with the carer 1319 so that the medical device 1316 is placed by the responding carer It can emit alerts (such as tones, flashing lights or electronic notifications). System and method for determining the position of a medical device close to a patient is a system for determining the position of a medical device relative to a wearable device, filed Dec. 18, 2015 and published as WO / 2016/106132 and Methods are described in International Patent Application No. PCT / US15 / 66720 entitled Invention. The application is hereby incorporated by reference in its entirety.

  In the example position 1320, the motion sequence of the patient 102 is classified as walking to the toilet 1330. At the exemplary location 1322, the motion sequence of the patient 102 is classified as walking out of the assigned room door. As mentioned above, the medical device 101 is worn by the patient 102. The motion sensor / orientation sensor of the medical device 101 records motion data and transmits this data to the remote server 1302 through the network 1306. At remote server 1302, motion recognition component or motion classification component 316 classifies this data using the various motion recognition algorithms described above. As mentioned above, depending on the threshold associated with the patient's 102 movement, an alert, alert or other action may be taken.

  In yet another example, the above-described example may be combined with, in combination with all of them, or any of them in order to track the position of the patient 102 in a hospital or clinic Not, it can be used in a hospital or clinical setting. Thus, in addition to monitoring patient movement and orientation, some examples are for proximity sensors, beacons, and various area networks (eg, Wi-Fi® or ZigBee® networks) Use the infrastructure to identify the location (or an approximate location depending on the measurement distinction of the technology used). Similar to some of the above examples, various thresholds regarding the location of the patient may be established so that the patient and / or caregiver is notified when the patient exceeds the threshold. Additional details for the example including tracking the position of the patient are described in US Patent No. 15 / 077,995.

  Having described several aspects of at least one example, it should be understood that various changes, modifications, and improvements will readily occur to those skilled in the art. For example, the examples disclosed herein may also be used in other contexts. Such alterations, modifications, and improvements are intended to be part of the present disclosure, and are intended to be within the scope of the examples set forth herein. Accordingly, the foregoing description and drawings are by way of example only.

Claims (20)

A plurality of motion sensors located at one or more anatomical locations on a patient's body and configured to detect a plurality of motion parameters corresponding to movements of a region of the patient's body;
At least one processor communicatively coupled to the plurality of motion sensors;
Receiving the plurality of motion parameters corresponding to the movement of the region of the patient's body,
Storing the plurality of motion parameters in a data store;
Processing the plurality of motion parameters stored in the data store to determine the motion of the portion of the patient's body;
Classifying the movement of the portion of the patient's body into a type based on a plurality of predetermined movement sentence features including at least one predetermined movement primitive;
A portable medical device for monitoring patient movement, comprising: at least one processor configured to initiate one or more notification actions based on the typology. The plurality of motion parameters include at least one orientation parameter corresponding to the orientation of the region of the patient's body,
A portable medical device according to claim 1. The plurality of motion sensors include at least one of an accelerometer, a gyroscope, and a magnetometer.
A portable medical device according to claim 1 or 2. The plurality of predetermined motion sentence features may include a plurality of predetermined motion primitives, at least one predetermined motion modifier, at least one predetermined motion object, and at least one predetermined one. Selected motion sentences and at least one predetermined series of motion sentences,
The portable medical device according to any one of claims 1 to 3. The at least one processor may include first sensor data acquired from a first motion sensor attached to a first position, and second data acquired from a second motion sensor attached to a second position. Are configured to classify the movement of the patient with reference to sensor data of
The portable medical device according to any one of claims 1 to 4. The first position is an anatomical position on the patient, and the second position is a position of a physical object other than the patient.
A portable medical device according to claim 5. The physical objects include at least one of a bed and a wheelchair
A portable medical device according to claim 6. The one or more anatomical locations on the patient's body include the patient's body's head, chest, leg, neck, shoulder, elbow, knee, wrist, jaw, forearm, biceps, ankles and feet Including one or more of
The portable medical device according to any one of claims 1 to 7. The at least one processor is configured to classify the motion based on one or more motion detection rules derived from a database of pre-collected motion information.
9. A portable medical device according to any one of the preceding claims. The at least one processor is
Classify the motion using motion recognition processing;
Training the motion recognition process using at least one of a predetermined movement measured from a plurality of patients and a patient specific movement derived during a baseline period;
10. A portable medical device according to any one of the preceding claims. The patient-specific movement derived during the baseline period is recorded during sleep and / or during a six-minute walk test period.
A portable medical device according to claim 10. The one or more notification actions include notifying a caregiver of the movement of the patient.
A portable medical device according to any one of the preceding claims. The one or more notification actions include alerting the patient based on the movement of the patient.
13. A portable medical device according to any one of the preceding claims. The at least one processor is
Identify another medical device remote from the portable medical device in proximity to the patient;
Configured to instruct the medical device to emit an alert,
14. A portable medical device according to any one of the preceding claims. The at least one processor is configured to identify a current time and to perform an action associated with the current time.
15. A portable medical device according to any one of the preceding claims. The at least one processor to classify the movement of the patient as at least one of a loss of directional movement, a fall, a fall, a whine, a fainting, a seizure, a tremor, and a cough. Configured,
16. A portable medical device according to any one of the preceding claims. With wearable defibrillator,
A portable medical device according to any one of the preceding claims. With a mobile cardiac monitoring device
18. A portable medical device according to any one of the preceding claims. Equipped with a heart monitoring device,
The cardiac monitoring device
A plurality of motion sensors located at one or more anatomical locations on the patient's body and configured to detect motion parameters corresponding to movements of a region of the patient's body;
At least one processor communicatively coupled to the plurality of motion sensors;
At least one processor configured to transmit the plurality of motion parameters corresponding to the movement of the region of the patient's body to a remote server;
The remote server is
Receiving the plurality of motion parameters corresponding to the movement of the region of the patient's body,
Storing the plurality of motion parameters in a data store;
Processing the plurality of motion parameters stored in the data store to determine the motion of the portion of the patient's body;
Classifying the movement of the portion of the patient's body into a type based on a plurality of predetermined movement sentence features including at least one predetermined movement primitive;
Configured to initiate one or more notification actions based on the type
System for monitoring patient movement. At least one client device in communication with the remote server, configured to notify at least one of a caregiver and the patient about the movement of the patient based on the one or more notification actions Further comprising at least one client device
A system for monitoring patient movement according to claim 19.